Insurance
An assistant answers claims handlers from the policy terms and internal procedures, and shows the article it used.
Generative AI · Companies, public bodies and creators
Generative AI: assistants on your data, visuals and videos for your campaigns
We build assistants that answer from your documents and cite their sources. We also produce your advertising visuals and videos, and automate your social media content. Your team approves everything before it goes live.
Time lost searching, and AI used without rules
A case handler hunts for the right clause through hundreds of pages of policy terms, circulars and internal notes.
Your teams paste extracts from client files into ChatGPT, with no rules and no idea where the data goes.
Your customers and citizens ask the same questions by phone, email and WhatsApp, and your advisers repeat the same answers.
Every week, your teams rewrite the same letters and make the visuals for your social media by hand.
A first chatbot gave a customer a made-up answer, and nobody could trace where it came from.
What you get
Your assistant cites its sources, and your team approves every visual and video before it goes live.
An internal assistant on your documents
It answers your teams from your procedures and contracts, cites the passage it used and prepares first drafts of your letters.
A chatbot or AI agent for your customers and citizens
On your website or on WhatsApp Business, it answers common questions and passes other requests to your team.
Automated content production
For creators and brands, a pipeline prepares scripts, visuals and short videos, then schedules each post once you approve it.
Advertising visuals and videos
We produce them for your campaigns and social media, within your brand guidelines. You approve each version before release.
Quality tests and answer rules
We test the assistant on questions your experts have approved. For out-of-scope questions, it hands over to a person.
Use cases by sector
Retail and distribution
A retailer uses AI to produce the visuals and videos for its campaigns, adapted to each social network. The marketing team approves each version.
Content creators
An automated pipeline prepares a creator’s scripts, visuals and short videos each week, then schedules the ones the creator has approved.
How we work
Use case workshop
We choose the use case with you: an assistant on your documents or content to produce. We set aside uses that carry too much risk.
You receive
the chosen use case, the document set or brand guidelines, and the quality criteria.
A first prototype
We build the assistant on a sample of your documents, or a first batch of content. Your teams test the result.
You receive
a working prototype and a report on its tests.
Rollout
We connect the assistant to your website or WhatsApp Business and the content pipeline to your social accounts, with logging and human handover.
You receive
a live tool and its documentation.
Quality monitoring
We review answers picked at random, fix errors and update the document set when your documents change.
You receive
a dashboard tracking quality and costs.
Data and safeguards
The assistant cites the document and passage it used. When it finds no source in your documents, it says so and refers the user to someone on your team.
A person on your team reviews each letter, report, visual or video the AI prepares before it is sent or published, and approves any sensitive action by an AI agent.
We choose the model and the hosting with you, in Morocco, on your servers or in the cloud, and we apply Moroccan law 09-08. The assistant follows your access rights: each user can query only the documents they can already open.
We log each question, answer and source. Before rollout, we estimate the monthly model usage cost, then compare actual spend with that estimate each month.
Technologies we use
- Language modelsGPT, Claude, Gemini
- Open modelsLlama, Qwen
- RAG and semantic search
- PostgreSQL and pgvector
- Python
- Next.js and TypeScript
- WhatsApp Business API
- Image and video generation models
Frequently asked questions
Generative artificial intelligence produces text, images, videos or code from a written instruction. A company uses it to answer questions about its documents, draft letters or produce advertising visuals and videos. A person checks the output before anyone uses it.
With RAG (retrieval-augmented generation), the assistant first searches your documents for the relevant passages, then the model answers from those passages alone and cites its sources. To update what it knows, you update the documents. Nobody retrains the model.
Yes, for creators and brands alike. We build a pipeline that prepares your scripts, visuals and short videos within your brand guidelines, then schedules them for publication. A person on your team approves each piece before it goes live.
A model can get things wrong. To reduce that risk, the assistant answers from your documents, cites the passage it used and declines out-of-scope questions. Before rollout, we measure its error rate, and we log each exchange so we can trace a wrong answer.
You decide with us where your data goes. For sensitive documents, we offer an open model hosted in Morocco or on your servers. If a model processes personal data outside Morocco, we help you prepare the filings with the CNDP, Morocco’s data protection authority. We apply law 09-08.
The price depends on the number of documents or the volume of content, the channels to open (website, intranet, WhatsApp, social networks) and the tools to connect. Before the project starts, you receive a quote for a written scope that lists the project price and the monthly model cost separately.
AI solutions delivered fast, built to last.
Let’s talk about your project
Describe your need in a few lines: the process involved and the tools you use today. For a public tender, include its reference. We reply by email or by phone.
Based in Rabat, we work across Morocco, Africa, the Middle East and Europe, on site or remotely.